Back

GAN-based anomaly detection in multi-modal MRI images

Sean Benson; Regina Beets-Tan

2020-07-12 bioinformatics
10.1101/2020.07.10.197087 bioRxiv
Show abstract

Generative adversarial networks (GANs) are known to be a powerful tool in order to correct image aberrations, and even predict entirely synthetic images. We describe and demonstrate a method to use GANs trained from multi-modal magnetic resonance images as a 3-channel input. The training of the generative network was performed using only healthy images together with pseudo-random irregular masks. The dataset consisted of just 20 people. The resulting model was then used to detect anomalies real patient images in which the anomaly was a tumour. The search was performed using no prior knowledge of the tumour location, if indeed a tumour was present. Resulting accuracies are observed to vary significantly on the size of the anomaly. The area under the receiver operator characteristic curve is observed to be greater than 0.75 for anomaly sizes greater than 4 cm2.

Matching journals

The top 9 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.